The modulatory properties of Astragalus membranaceus treatment on endometrial cancer: an integrated pharmacological method

Data mining based on the GEO and TCGA database

The gene expression profile GSE63678 dataset (Pappa et al., 2015) was downloaded from the GEO database (https://www.ncbi.nlm.nih.gov/geo/). The platform for GSE63678 is GPL571 ([HG-U133A_2] Affymetrix Human Genome U133A 2.0 Array), including five normal tissues and seven tumour tissues. The probes were converted into the corresponding gene site according to the annotation information on the platform. All gene expression data were subjected to log2 transformation.

The Cancer Genome Atlas (TCGA) is a free and comprehensive database that includes large clusters of human cancer genome sequencing data (https://cancergenome.nih.gov/). The standard for data inclusion in our study was that the sample size must consist of both cancer and normal tissues, and that the number of normal tissue controls must be greater than or equal to three. In this study, we obtained 587 cases of gene expression data (35 cases of normal endometrial samples and 552 cases of EC samples) from TCGA official website for the Uterine Corpus Endometrial Carcinoma projects (UCEC). Before further analysis and processing, RNA sequencing data standardization was carried out in full accordance with TCGA’s release guidelines.

Identification of DEGs in EC

The “affy” R language package was applied to complete the log2 transformation, quantile normalization and median polish algorithm summarization. The probes were annotated by the Affymetrix annotation files. Microarray quality was evaluated by sample clustering based on the distance between different samples in Pearson’s correlation matrices (Liu et al., 2019b). The limma package was used to detect DEGs between the normal and EC tissues (Zhang et al., 2019). Screening criteria of —log2-fold-change (FC)— ≥ 1 and P-value < 0.05 were considered indicators of statistical significance in the identification of robust DEGs. In the same way, we used the R software (version: x64 4.1.0) and the “edgeR” package to screen the DEGs in the TCGA dataset. The screening criteria were: —log2-fold-change (FC)— ≥ 1 and P-value <0.05. Subsequently, we took the intersected DEG s of the two datasets by the Venn package in the R software.

Chemical compound database building

The Traditional Chinese Medicine System Pharmacology Database (Ru et al., 2014) (TCMSP™, https://tcmsp-e.com/) is a resource of systems pharmacology for TCMs and related compounds. We collected the data on Astragalus membranaceus active compounds using the ADME (absorption, distribution, metabolism, and excretion) filter method, whose main parameters included oral bioavailability (OB) and drug-likeness (DL) (Zhang et al., 2016). OB of ≥30% and DL of ≥0.18 were the threshold values for ADME evaluation systems. When herbal compounds met both criteria (OB ≥ 30%, DL ≥ 0.18), they were selected as candidates for further analysis. The chemical name “Astragalus membranaceus” was entered into the search box, and its pharmacokinetic properties were studied at the molecular level.

Disease targets database building and network construction

We introduced Astragalus membranaceus-regulated target genes and EC-related target genes into the Perl script to get overlapping targets and active compounds, which were the potential targets of Astragalus membranaceus against EC. Subsequently, these overlapping parts were introduced into Cytoscape 3.7.1 software (http://www.cytoscape.org/) (for network construction and visualization) (Holmås et al., 2019). In the network, “node” represents the corresponding compound and target genes, and “edge” represents the relationship between the compound and the target genes.

Protein–protein interaction (PPI) network construction and topology analysis

We introduced the overlapping gene targets into Cytoscape software for PPI analysis. The BisoGenet 3.0.0 package in the Cytoscape software was used to construct a protein interaction network (Martin et al., 2010). Then, we used the “CytoNCA” plugin of the Cytoscape software for topology analysis (Tang et al., 2015). For each node in the interaction network, “degree (DC)” and “betweenness centrality (BC)” were selected to assess its topological features. DC represents the number of node links and reflects how often a node interacts with other nodes (Missiuro et al., 2009). BC is defined as the degree to which a node is located on a path between other nodes (Raman, Damaraju & Joshi, 2014). The original PPI network was screened twice according to the values of DC and BC to obtain the core PPI network.

Gene function and pathway enrichment analysis

The “ggplot2” package in R software for Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis (Maag, 2018), to obtain the biological process, cellular compounds, molecular function, and critical signalling pathways (The Gene Ontology Consortium, 2017). In addition, to better understand the complex relationship between core biological pathways and hub co-targets associated with the Astragalus membranaceus, we used Cytoscape 3.7.1 (http://www.cytoscape.org/) to build and analyse networks.

Reagents and cells

Formononetin (C₁₆H₁₂O₄, 98% purity verified by high-performance liquid chromatography) was obtained from Phytomarker Ltd., Tianjin, China. Formononetin was dissolved in dimethyl sulfoxide (DMSO) as a stock solution of 100 mM. The 40 µM and 80 µM concentrations of formononetin in experiments were obtained by diluting the medium to a final concentration of DMSO< 0.1%(vol/vol). Cells treated with 0.1% DMSO served as a control. The solution was stored at 4 °C for further use.

Human endometrial cancer cell lines (Ishikawa, HEC-1A, and HEC-251) were purchased from the Institute of Biochemistry and Cell Biology of the Chinese Academy of Sciences (Shanghai, PR China). All cell lines were maintained in Dulbecco’s modified Eagle medium (DMEM), or DMEM/F12 supplemented with 10% heat-inactivated fetal bovine serum (FBS) and kept at a 37 °C, 5% CO₂ incubator. Cell lines were routinely checked for mycoplasma contamination.

Cell viability assay

Cell proliferation was performed by a Cell Counting Kit-8 (CCK-8) (Meilun Biotechnology Co., Ltd., MA0218-5) according to the manufacturer’s protocol. Briefly, the Ishikawa, HEC-1A, and HEC-251 cell lines were trypsinized and seeded into 96-well plates at a density of 1 ×10⁴ each well. After incubating overnight at 37 °C, the cells were treated with a concentration of 40 µM and 80 µM of formononetin for 24 h. The absorbance of each well was determined at 490 nm under a Smart Microplate Reader (USCNK, Wuhan).

Colony formation assay

The effect of formononetin on the colony-forming ability of endometrial cancer cells was evaluated by colony formation assay. The Ishikawa, HEC-1A, and HEC-251 cell lines were seeded into 6-well plates at the concentration of 400 cells per well. After incubating for 24 h at 37 °C, the cells were incubated with formononetin (40 µM and 80 µM). Three days later, most single colonies contain more than 50 cells. Cells were harvested and rinsed three times with distilled water, and images were obtained by a microscope.

RNA extraction and qPCR assays

After exposure to formononetin (40 µM and 80 µM) for 48 h, the Ishikawa, HEC-1A, and HEC-251 cells were harvested for RNA isolation by using a TRIZOL^®reagent (TaKaRa, 9109). Then, cDNA was subsequently synthesized using 10 ng of RNA under standard conditions for the iScript cDNA Synthesis Kit (Fermentas Inc., USA). Quantitative polymerase chain reaction (qPCR) was performed by SYBR Premix Ex Taq II (Takara, Japan). The 2^−ΔΔCt method was used to calculate the relative expression. The gene primers were listed as follows: ERβ: 5′-AATGGGGTTCTCTCCTGT’ (sense) and 5′-AGCCCAAAGTATCCCTGAC-3′(antisense); p53: 5′-CAGTAGTCAAGTAGTAACCCCTGCCTTGCACAG-3′ (sense) and 5′-CATGTAT-TACTGTGCAAGGCAGGGGTTACTACTT-3′ (antisense); Actin: 5′-CATGTACGTTGCTATCCAGGC-3′(forward) and 5′-CTCCTTAATGTCACGCACGAT-3′(reverse).

Western blotting

After 48 h of incubation with the designated concentration of formononetin, the Ishikawa, HEC-1A, and HEC-251 cells were lysed in RIPA buffer. The quantified protein was separated by 10% SDS-PAGE and subsequently transferred to polyvinylidene difluoride (PVDF) membranes (Bio-Rad, USA). Then the membrane was sealed with 5% nonfat milk and incubated with primary antibodies including ERβ (1:2500; Proteintech Group, Inc. Wuhan, China), p53 (1:800; bsm-33058M; Beijing Bioss Biological Co., Ltd, Beijing, China), and β-actin (ABclonal, Inc. Wuhan, China) for 2 h at room temperature. After three washes with the TBST solution, the samples were incubated with the appropriate secondary antibodies for 2 h. The blots were visualized with an electro-chemiluminescence (ECL) reagent kit (Beyotime, China). Image J software (National Institutes of Health, Bethesda, MD, USA) was used to quantify each protein band intensity, and the β-actin was used as an internal reference.

Statistical analysis

The data were presented as the mean ± standard deviation (SD). The SPSS 25.0 software (SPSS, Chicago, IL) was used for statistical analysis using one-way ANOVA. All experiments were carried out at least three times, and a P < 0.05 was considered significant.

Endometrial cancer-related targets

Based on the DEG selection criteria of —log2-fold-change (FC)— ≥ 1 and P-value < 0.05, a total of 360 DEGs, including 204 upregulated DEGs and 156 downregulated DEGs, were identified between the normal and EC tissues after consolidation and normalization of the microarray data from the GSE63678 dataset (Fig. 1). In addition, a total of 11,531 DEGs were obtained from the TCGA database. Cluster analysis (Fig. 2) revealed 7,727 upregulated genes and 3,804 downregulated genes. The Venn package was used to screen the intersecting DEGs from both databases and generate the Venn map (Fig. 3). Finally, 268 EC-related DEGs with high reliability were obtained (Table S1 for details).

Active compounds in Astragalus membranaceus and the targets

The effective compounds of Astragalus membranaceus were screened in the TCMSP database. A total of 87 chemical constituents and 953 targets of Astragalus membranaceus were searched, and 20 potentially active compounds, screened by OB > 30% and DL > 0.18, and 463 related targets of Astragalus membranaceus were selected. Table 1 shows the information on the top 10 active compounds and their targets (Table S2).

Disease targets database building and network construction

A total of 10 overlapping genes and 8 overlapping potentially active compounds were screened as all targets for Astragalus membranaceus treatment of EC (Table 2). As shown in Fig. 4, the compound-target network consisted of 18 nodes and 19 edges. The top three compounds of Astragalus membranaceus with the highest values of both DC and BC, in descending order by DC, were kaempferol (MOL000422), formononetin (MOL000392), and quercetin (MOL000098), and they will likely play a more significant role in treating EC.

PPI network construction and topology analysis

We obtained the core PPI network after the original PPI network screened according to DC and BC (Fig. 5). The core PPI network consisted of 48 nodes and 528 edges. In other words, the core PPI network contained 48 compounds and target genes, and there were 528 relationships between them (Fig. 5C). Genes, namely, TP53, NPM1, YWHAZ, HSP90AA1, and BRCA1 with the highest DC and BC values, are likely the critical targets in EC.

Biological functions and pathway enrichment assays of the core targets of Astragalus membranaceus and EC

GO analysis (Figs. 6A–6C) and pathway enrichment analysis (Fig. 6D) were performed to clarify the biological functions of these 10 target genes. The screening condition was P < 0.05 and Q < 0.05. The BP category of the GO analysis results showed that the target genes were markedly associated with histone phosphorylation, DNA integrity checkpoint, and signal transduction involved in DNA integrity checkpoint. For the CC category, the target genes were markedly related to the cyclin-dependent protein kinase holoenzyme complex, condensed chromosome and serine/threonine protein kinase complex. The MF category was significantly enriched in histone kinase activity, heat shock protein binding, and Hsp70 protein binding. The KEGG signalling pathway analysis showed that the target genes played pivotal roles in the following pathways: ‘p53 signalling pathway’, ‘Transcriptional misregulation in cancer’, ‘EGFR tyrosine kinase inhibitor resistance’, and ‘Endometrial cancer’.

In addition, based on the target and pathway analyses, an entire compound, target and disease network was constructed by Cytoscape (v 3.7.1). As shown in Fig. 7, this compound, target, and disease interaction network had 23 nodes and 43 edges. Red inverted triangles and orange rectangles correspond to the pathway and target genes, respectively.

Anti-proliferation effect of formononetin on endometrial cancer cells

Formononetin, as an immunomodulator, has been widely used for centuries. Formononetin’s anti-tumour effects such as inhibiting cell proliferation, inducing cell cycle arrest and apoptosis have been confirmed in bladder cancer (Wu et al., 2017), breast cancer (Xin et al., 2019), and ovarian cancer (Park et al., 2018). However, the role of formononetin in endometrial cancer cells remains to be explored. Therefore, we choose formononetin for verification.

The effect of formononetin on the proliferation of endometrial cancer cell lines was tested by cell viability assay and clone formation assay. As shown in Fig. 8, formononetin significantly inhibited the proliferation rate of endometrial cancer cell lines compared with the control group (all P < 0.05 relative to the control group). In addition, compared with 40 µM of formononetin, higher concentrations inhibited the cells more strongly (all P < 0.05 relative to the 40 µM group). These results indicate that formononetin may inhibit cell proliferation in a dose-dependent manner.

The effect of formononetin on the expression of ERβ and p53

ERβ has been widely reported to play an important role in inhibiting cell proliferation. Furthermore, KEGG pathway analysis showed that Astragalus membranaceus could act on EC through the p53 signalling pathway. In order to explore how formononetin exerts an anti-tumor effect, we explore the differences in ERβ and p53 between different groups. In different groups of endometrial cancer cell lines, the expression of ERβ and p53 in the cells treated with formononetin was significantly increased, as expected (all P < 0.05 relative to the control group); In addition, the higher the concentration of formononetin, the more the expression of ERβ and p53 (all P < 0.05 relative to the 40 µM group; Figs. 9A and 9B). To add more evidence, we performed Western blots on ERβ and p53 using the same cells. Indeed, formononetin can promote the expression of ER β and p53 in a dose-dependent manner (Fig. 9C).